Mapping the change: study of semantic change segmentation using mobile 3D point clouds
ABDERRAZZAQ KHARROUBI | In Progress
Summary
Our world is in a state of continuous change, making it essential to monitor and understand these changes effectively. My research project focuses on detecting changes by analyzing the semantics of objects, providing deeper insights into the dynamics involved. To achieve this, I leverage recent advancements in 3D semantic segmentation, utilizing kernel convolution and transformer techniques to enhance detection accuracy.
A key element of this research is the use of 3D point clouds, which are crucial for accurate change detection. 3D point clouds offer detailed spatial information about the environment, capturing the precise shape and position of objects. This high-resolution data is invaluable for detecting subtle changes that might be missed with traditional 2D methods. By using mobile 3D point clouds, we can dynamically and comprehensively survey environments, making it possible to monitor changes in both urban and railway scenes.
The project is divided into three main components. The first component involves assessing semantic segmentation to ensure precise classification of objects within the 3D point clouds. This step is crucial as it lays the groundwork for accurate change detection. The second component is the development of an unsupervised method for change detection, which allows for the identification of changes without the need for extensive labeled data. This method leverages occupancy grids and decision trees to make informed decisions about changes.
The third component of the research focuses on propagating uncertainty to provide a metric for the relevance of detected changes. By accounting for uncertainty, we can better understand the significance of changes and improve the reliability of our findings. This aspect is particularly important in dynamic environments where changes can have varying levels of impact.
Overall, this research aims to improve our ability to understand changes in the environment by integrating advanced semantic segmentation techniques with robust change detection methods. The use of 3D point clouds enhances our ability to capture detailed spatial information, leading to better-informed decisions and more effective management of urban and railway areas.
Comment
This doctoral research is funded by the Belgian National Funds for Scientific Research FNRS, by an Aspirant grant for Abderrazzaq Kharroubi.
Links
- Kharroubi, A., Ballouch, Z., Hajji, R., Yarroudh, A., & Billen, R. (09 April 2024). Multi-Context Point Cloud Dataset and Machine Learning for Railway Semantic Segmentation. Infrastructures, 9 (4), 71. doi:10.3390/infrastructures9040071 https://hdl.handle.net/2268/316233
- Tamort, A., Kharroubi, A., Hajji, R., & Billen, R. (08 March 2024). 3D CHANGE DETECTION FOR SEMI-AUTOMATIC UPDATE OF BUILDINGS IN 3D CITY MODELS. International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLVIII-4/W9-2024, 349-355. doi:10.5194/isprs-archives-xlviii-4-w9-2024-349-2024 https://hdl.handle.net/2268/314264
- Kharroubi, A., Poux, F., Ballouch, Z., Hajji, R., & Billen, R. (17 October 2022). Three Dimensional Change Detection Using Point Clouds: A Review. Geomatics, 2 (4), 457-486. doi:10.3390/geomatics2040025 https://hdl.handle.net/2268/296036
updated on 6/10/24
